1234

OTC-27799-MS

Quantification and Modelling of the Dynamic Wake Effect for Floating

Offshore Wind Turbines

Abdulqadir Aziz Singapore Wala, Lloyd's Register Global Technology Centre; Eddie Yin-Kwee Ng, Nanyang
Technological University; Anand Bahuguni, Lloyd's Register Global Technology Centre; Narasimalu Srikanth,
Nanyang Technological University

Copyright 2017, Offshore Technology Conference

This paper was prepared for presentation at the Offshore Technology Conference held in Houston, Texas, USA, 1–4 May 2017.

This paper was selected for presentation by an OTC program committee following review of information contained in an abstract submitted by the author(s). Contents of
the paper have not been reviewed by the Offshore Technology Conference and are subject to correction by the author(s). The material does not necessarily reflect any
position of the Offshore Technology Conference, its officers, or members. Electronic reproduction, distribution, or storage of any part of this paper without the written
consent of the Offshore Technology Conference is prohibited. Permission to reproduce in print is restricted to an abstract of not more than 300 words; illustrations may
not be copied. The abstract must contain conspicuous acknowledgment of OTC copyright.

Abstract
Modelling the aerodynamic forces on floating offshore wind turbines (FOWTs) is a challenging task due to
the motion of the floating platform, which result in flow transients and associated aerodynamic effects. Each
of these needs to be modelled and implemented, before a complete aerodynamic model of the FOWT can be
presented. Of special interest is the dynamic wake effect, a result of the time lag between rotor forces and the
air flow deceleration within the wake. The dynamic wake effect may present itself significantly in several
scenarios including gust loads, changing wind speeds and direction, and the oscillatory motion of the rotor
due to platform motion from wave forces. The complexity of the problem is substantiated by the added mass
effect, which must be accounted for when considering rotor motions through the air. Computational fluid
dynamics (CFD) simulations and blade element momentum (BEM) computations have been performed for
the NREL 5MW virtual wind turbine in axial flow and surge motions at different frequencies and amplitudes,
to quantify and model the dynamic wake effect, including the added mass effect for rotor motion. This
dynamic wake model is then implemented in an unsteady blade element momentum (uBEM) code, for a
more complete model of wind turbine aerodynamics.

Introduction
Wind turbine aerodynamics is composed of a balance between rotor aerodynamic forces and wake
momentum balance. General momentum theory can be used to resolve this balance, and by means of an
appropriate local aerodynamics model, obtain the force distributions along a wind turbine blade. In fixed-
bottom wind turbines in steady or pseudo-steady wind conditions, the blade element momentum (BEM)
method can accurately predict blade elemental forces based on flow conditions. In floating offshore wind
turbines (FOWTs), where the platform motions can be expected to give rise to rotor-level flow transients that
result in a constantly-changing momentum balance, a time-lag is expected between the rotor aerodynamic
forces and momentum balance, since the flowfield momentum changes arising from the forces would take
time to reach the far-wake.
2 OTC-27799-MS

Studies on the Tjæreborg 2MW wind turbine [1] have shown that pitch changes result in in a spike in
aerodynamic forces, before gradually returning to steady-state values as predicted by momentum theory.
This is due to the changes in inflow resulting in changes in rotor aerodynamic forces, resulting in changes
in the far-wake induced velocities, in that sequential order.
In order to account for this time-lag, the momentum balance may either be obtained through unsteady
Euler equations, such as the Pitt and Peters model [2] and the generalized dynamic wake model [3], or by
correcting the 1-dimensional momentum equations through a dynamic wake model. Single and 2-differential
equation dynamic wake models have been proposed in [4], including the ECN model and Øye model [5].
The Øye model used two differential equations with two time constants to account for near-wake and far-
wake time scales.
The wake flow field influences the induced velocity. This affects the angle of attack of the rotor blade. In
a steady wind, this angle of attack would be constant, since the rotor aerodynamic forces would be steady,
resulting in a steady effect on the flow field, and thus a steady influence on the rotor aerodynamics. In
unsteady flow conditions, there are two effects to be considered:
1. The angle of attack experienced by the rotor blade cross-sections, thus aerodynamic force distribution
is dependent on the wake velocities.
2. The wake flow is dependent on the force distribution on the rotor blades, which is dependent on the
rotor aerodynamic forces.
The first part refers to the unsteady airfoil effects experienced by the blade cross-sections, while the
second part describes the dynamic wake effect. Due to the interconnected nature of the two effects, it is often
difficult to differentiate when changes in the rotor aerodynamic forces and the wake are due to the dynamic
wake effect and when they are due to the unsteady airfoil effect. This can be resolved by modelling the
unsteady airfoil effect based on a 2D understanding of the effect, corrected for stall-delay. Having applied
the appropriate model to an unsteady blade element momentum model, further time lags between forces
and wake velocities can be assumed to be due to the dynamic wake effect.
Based on the differences between calculated forces in an unsteady blade element momentum algorithm
and rotor forces extracted from CFD, the time parameters in the 2-equation Øye model may be optimized
to match the CFD-extracted forces, thus provide a clearer understanding of the magnitude and time scales
of the dynamic wake effect. The present study was conducted using an analysis of the NREL 5 MW virtual
turbine [6] in axial flow and surging motion.

Methodology
Simulation Matrix
A total of 9 simulations were conducted, with sinusoidal surge motions at 3 different frequencies and 3
different amplitudes. The simulated wind speed was 10m/s at 11.45RPM. The surge motion frequencies and
amplitudes are summarized in Table 1.

Table 1—Sinusoidal surge motion amplitudes and frequencies in present study

Amplitudes (m) Frequencies (Hz)

2, 4, 6 0.02, 0.10, 0.50

Computational Fluid Dynamics Methodology

The CFD simulations were performed in ANSYS CFX 16. A single blade was used in 120° periodic
simulations to reduce computational time. A frozen rotor approach was employed to simulate the rotation
of the blade. The k-ω SST turbulence model was used with 5% turbulence intensity. The fluid was air at
OTC-27799-MS 3

25°C. A high resolution scheme was used for advection and turbulence numerics, while a second-order back
Euler scheme was used for time-stepping. A fourth order Rhie-Chow interpolation was used for pressure-
velocity coupling.
The inlet boundary was a velocity inlet with uniform velocity of 10m/s, and the outlet was a pressure
outlet. The side boundaries were set as openings, while the blade was a no-slip wall.
Mesh Setup. The mesh was divided into 3 regions: Blade, Near-wake and Far-wake. The blade region
was situated directly around the blade, in an o-grid. The height of the first layer of the blade was 6.6μm to
8.2μm, depending on position along the blade. This is shown in Figure 1.

Figure 1—O-grid generated around the rotor blade

The mesh was refined to ensure that y+<1 over most of the blade, as shown in Figure 2. The y+ only
exceeds 1 near the tip, but is still low at y+=3.

Figure 2—y+ over blade during static simulation

The near-wake region is a cylindrical region of refined mesh to ensure that the flow features of the near-
wake a satisfactorily resolved for the accurate determination of blade forces. The near-wake mesh region
extends 1R, or 1 rotor -radius upstream and 3.33R downstream of the rotor. The radius of the region is
1.17R. The near wake and blade regions were set to rotate at 11.45 RPM in the frozen rotor approach.
The far-wake region extends 4.8R upstream and 11R downstream of the rotor. Its radius is 6R. The near-
wake and far-wake regions are shown in Figure 3 and Figure 4.
4 OTC-27799-MS

Figure 3—Front view of computational domain

Figure 4—Side view of computational domain

A deforming mesh was used to simulate the rotor surge motion. The near-wake and blade regions were
given a high mesh rigidity, thus deformation was limited to the far-wake region, to prevent negative volumes.
The surge motion was defined by the equation:
(1)
where S is the displacement of the rotor, A is the amplitude of motion and f is the frequency of motion.
Mesh Refinement. A mesh refinement study was carried out with a coarse (3.3 million cells), medium,
(3.7 million cells) and fine (6 million cells) mesh. The mesh refinement was carried out on a static rotor,
then on 2 cases of sinusoidal surge motions with amplitudes and frequencies shown in Table 2.

Table 2—Sinusoidal surge conditions in present study

Amplitude (m) Frequency (Hz)

2 0.02
4 0.10

The percentage differences in elemental thrust and shaft torque from the medium mesh, for the coarse
and fine meshes, in the static case, are recorded in Table 3.
OTC-27799-MS 5

Table 3—Percentage change in thrust and shaft torque at each blade segment, compared to medium mesh for static simulation

Thrust (%) Torque(%)

r/R coarse fine coarse fine

0.19 3.02 1.05 2.18 0.81

0.25 0.29 0.14 0.28 0.15
0.32 0.01 0.00 0.02 0.05
0.38 0.03 0.06 0.04 0.08
0.45 0.00 0.09 0.03 0.10
0.51 0.00 0.00 0.01 0.10
0.58 0.00 0.07 0.00 0.08
0.64 0.00 0.00 0.01 0.07
0.71 0.00 0.00 0.01 0.09
0.77 0.00 0.00 0.00 0.10
0.84 0.00 0.05 0.00 0.09
0.89 0.00 0.00 0.04 0.06
0.93 0.00 0.00 0.02 0.08
0.98 0.01 0.02 0.02 0.07

The forces at each element had a maximum difference of 1.05% for thrust and 0.81% for torque between
the medium and fine meshes. Thus, the medium mesh was deemed sufficiently refined for the static rotor.
The time step for the dynamic simulations was ΔT = 1/(200f), meaning 200 timesteps per oscillatory
period. The percentage error in rotor thrust and shaft torque from the medium mesh was plotted against time
for the coarse and fine meshes in for the rotor in sinusoidal surge motion in Figure 5 to Figure 8.

Figure 5—Percentage error in rotor thrust relative to medium mesh for

sinusoidal surge motion with Amplitude=2m and Frequency=0.02Hz
6 OTC-27799-MS

Figure 6—Percentage error in rotor shaft torque relative to medium mesh

for sinusoidal surge motion with Amplitude=2m and Frequency=0.02Hz

Figure 7—Percentage error in rotor thrust relative to medium mesh

for sinusoidal surge motion with Amplitude=4m and Frequency=0.1Hz

Figure 8—Percentage error in rotor shaft torque relative to medium mesh

for sinusoidal surge motion with Amplitude=4m and Frequency=0.1Hz

All differences were less than 1% from the medium mesh, and not proportional to the surge motion. Thus,
the medium mesh was deemed acceptable for the study.
Time Step Refinement. A time step refinement study was conducted using the medium mesh. The fine
time step was half the size of the medium time step, at 400 time steps per period, and the coarse time step
was twice the size of the medium time step, at 100 time steps per period. The total thrust and shaft torque
against time of the coarse and fine time steps were compared to the medium time step, and the percentage
difference plotted in Figure 9 to Figure 12.
OTC-27799-MS 7

Figure 9—Percentage error in rotor thrust relative to medium time step

for sinusoidal surge motion with Amplitude=2m and Frequency=0.02Hz

Figure 10—Percentage error in rotor shaft torque relative to medium time

step for sinusoidal surge motion with Amplitude=2m and Frequency=0.02Hz

Figure 11—Percentage error in rotor thrust relative to medium time step

for sinusoidal surge motion with Amplitude=4m and Frequency=0.1Hz
8 OTC-27799-MS

Figure 12—Percentage error in rotor shaft torque relative to medium time

step for sinusoidal surge motion with Amplitude=4m and Frequency=0.1Hz

The results from Figure 9 to Figure 12 show that the coarse time step had more than 1% difference from
the medium time step, while the fine time step consistently had less than 1% difference. This suggests that
the medium time step is refined enough to capture the transient flow phenomena.

Unsteady Blade Element Momentum Method

The unsteady blade element momentum method (uBEM) used in the present study, is based on the method
presented in Hansen (cite).
The relative velocity vector is given by:

(2)

where V0 is the blade-specific wind speed at the blade element including relative speed due to rotor motion,
Ω is the rotor rotational speed in rads−1, r is the local blade segment radius.
The flow angle, ϕ, is the angle between the relative velocity and the normal of the rotor plane.

(3)

The normal (z) and tangential (y) induced velocities are then given by:

(4)

(5)

where ρ is the air density, B is the number of blades, F is the tip-loss factor, G is the Glauert correction for
high axial induction factors and Cl is the coefficient of lift corrected for stall-delay and dynamic stall at the
angle of attack equal to ϕ less the blade twist. Wx is set to 0.
At every time step, for each blade element in each blade, equations (2) to (5) are iterated until there is
no change in the induced velocity vector.
Dynamic Wake Model. Once the quasi-static value of the induced velocity is obtained, it is passed through
a dynamic wake model to account for the time-lag between changes in rotor forces and wake velocities. This
leads to an unsteady relative velocity which is then applied in the calculation the blade elemental forces.
OTC-27799-MS 9

The 2-differential equation Øye model was optimized in the present study for applications to FOWTs. The
Øye model is given by:

(6)

(7)

where Wqs is the quasi-static induced velocity vector obtained using equations (4) and (5), Wint is the
intermediate induced velocity vector obtained at the first time scale, and W is the dynamic wake induced
velocity. τ1 and τ2 are time scales given by:

(8)

(9)

where a is the axial induction factor given by:

(10)

This formulation is applied to account for possible changes in effective wind direction when rotor motion
is considered.
The right hand side of equation (6)) can be resolved using a backward-difference scheme, using which
the left hand side and equation (7)) may be solved analytically [7]. a cannot exceed 0.5 in this formulation.
Modified Dynamic Wake Model. The dynamic wake model of Øye was modified to increase the range of
applicability to all induced velocities and optimized to match rotor thrust and shaft torque obtained from
CFD of the NREL 5MW rotor in sinusoidal surge motion.
Within the modification, the differential equations remain the same, but the time constants were changed
to extend the range of applicability of the model and better reflect observed rotor thrust and shaft torque
from CFD.

(11)

(12)

The dynamic wake effect was observed to be minimal in the windmill and turbulent wake states, thus
τ1 was modified to be smaller within the windmill and turbulent wake states, but grow larger outside these
wake states.
Rotor forces. Re-applying the dynamic wake-corrected W to equations (2) and (3), the unsteady relative
velocity and flow angle is found. These are used to find the blade elemental thrust, T, and shaft torque, M:
(13)
(14)
where Cl and Cd are the coefficients of lift and drag respectively, corrected for stall-delay and dynamic stall
at the angle of attack equal to ϕ less the blade twist, and Δr is the blade element length.
The total of T and M for all elements of all blades will determine the rotor thrust and shaft torque.
10 OTC-27799-MS

Results and Discussion

The thrust and shaft torque against time are compared for a single blade of the NREL 5MW rotor in
sinusoidal surge motion in Figure 13 to Figure 18.

Figure 13—Rotor thrust for sinusoidal surge motion with amplitude=2m and frequency=0.02Hz

Figure 14—Rotor thrust for sinusoidal surge motion with amplitude=4m and frequency=0.1Hz

Figure 15—Rotor thrust for sinusoidal surge motion with amplitude=6m and frequency=0.5Hz
OTC-27799-MS 11

Figure 16—Rotor shaft torque for sinusoidal surge motion with amplitude=2m and frequency=0.02Hz

Figure 17—Rotor shaft torque for sinusoidal surge motion with amplitude=4m and frequency=0.1Hz

Figure 18—Rotor shaft torque for sinusoidal surge motion with amplitude=6m and frequency=0.5Hz

Figure 13, Figure 14, Figure 16 and Figure 17 show that the dynamic wake effect is insignificant at lower
motion frequencies. At these frequencies, the wind turbine operates within the windmill and turbulent wake
states. This justifies the modification to τ1 in equation (11)), which is designed to reduce the influence of
the dynamic wake model within the windmill and turbulent wake states. At high frequencies, the dynamic
wake model becomes more significant, as is demonstrated in the difference in rotor shaft torque prediction
with and without the dynamic wake model in Figure 18.
12 OTC-27799-MS

Conclusion
A dynamic wake model is presented based on modifications to the 2-differential equation Øye model. The
first time constant of the model was modified to improve the range of applicability of the model, and
optimized to match CFD-computed rotor thrust and shaft torque data.
Comparisons between the CFD-computed rotor thrust and shaft torque data and the unsteady BEM model
both with and without the dynamic wake model, demonstrated the differences between a dynamic wake
assumption and a quasi-static wake assumption. At high frequencies and axial induction factors beyond the
windmill and turbulent wake states, where FOWTs can be expected to operate, the dynamic wake effect was
shown to be highly significant, and modelling this effect was paramount to achieving a complete model of
wind turbine aerodynamics. A high level of accuracy was achieved with the proposed dynamic wake model.
The present study was performed in axial flow with surge motions, and future work should extend this
modelling approach to yawed flow cases, and more rotor models.

References
1. H. Snel and J. Schepers, "Investigation and modeling of dynamic inflow effects," in ECWEC ‘93
Conference, Travemünde, Germany, 1993.
2. D. Pitt and D. Peters, "Theoretical prediction of dynamic inflow derivaties," Vertica, vol. 5, no. 1,
pp. 21–34, 1981.
3. P. J. Moriarty and A. C. Hansen, "Aerodyn Theory Manual," NREL/TP-500-36881, Boulder,
Colarado, USA, 2005.
4. H. Snel and J. G. Schepers, "Joint Investigation of Dynamic Inflow Effects and Implementation
of an Engineering Method," ECN-C––94-107, Petten, The Netherlands, 1995.
5. S. Øye, "Unsteady wake effects caused by pitch angle changes," in Processdings of the first IEA
symposium on the aerodynamics of wind turbines, London, UK, 1986.
6. J. Jonkman, S. Butterfiled, W. Musial and J. Scott, "Definition of a 5-MW reference wind turbine
for offshore system development," Golden, Colarado, USA, 2009.
7. M. O. L. Hansen, Aerodynamics of Wind Turbines, 2nd ed., London, UK: Earthscan, 2008.